Atomic column elemental mapping is getting popular, since elemental species and positions of atomic sites are determined simultaneously [1]. In this method, the chemical information is generally detected by electron energy loss spectrometry (EELS) and/or energy dispersive X-ray spectrometry (EDS). Allowable electron dosage for a sample limits the usage of the method, since more dosage is required for the chemical analyses than that for imaging, due to small ionization cross section of atoms. The sensitivity of EDS rises rapidly, since a silicon drift detector (SDD), which is a new type of EDS detector, has design flexibility of its shape and has quick processing time, resulting in predominantly use in recent years. And total solid angle of X-ray detection rises rapidly to be 1.5-2.0 sr by detection systems with multiple detectors.

The allowable electron dosage still limits the application of the method to battery, carbon and organic materials, which are strongly requested to be analyzed by the industries. Therefore, it is required to reduce the dosage or to increase critical dosage of these materials. Trials to increase the critical dosage have been succeeded by finding an appropriate accelerating voltage and sample cooling. On the other hand, not so many trials to reduce the dose density onto a sample have been done.  We first succeeded in showing a pseudo atomic column elemental map with a lower average dose density ( < 1 % of that used in the conventional atomic column elemental mapping), utilizing a two dimensional (2D) moiré pattern [2].  In this paper, we applied this method to a beam sensitive sample.

The sample for our experiment was selected to be beryl (Be₃Al₂Si₆O₁₈: Fe²⁺) (known as aquamarine, having hexagonal structure with a = b = 0.922 nm, c = 0.920 nm see Fig. 1), which is one of cyclosilicates and has channels along the c axis. An [001] oriented sample was made by Ar ion thinning. Carbon was evaporated on the sample surface to avoid sample charging. We used for the experiment an aberration corrected 300 kV microscope (JEOL, JEM-ARM300F) equipped with a cold FEG and dual SDD X-ray detector system (total solid angle = 1.63 sr). All image observations and analyses in this paper were obtained under conditions: acc. Volt. = 300 kV, probe current = 24 pA. It is noteworthy that no direct atomic column elemental mapping was succeeded due to the sample damage. A high resolution STEM image of the sample is shown in Fig. 2(b).

In the experiments, the number of pixels (n x n) for the maps was selected to be 64 x 64, and pixel intervals in the x and y directions( d_rx and d_ry ) were set to be a_cr_x and b_cr_y nm, where r_x and r_y are numbers of unit cells in a pixel intervals in x and y directions.  The widths of the unit cell (a_c and b_c) are 0.922 and 1.60 nm, since b_c / a_c = 3^1/2 (see 2D Cartesian unit cell in Fig. 1(a)). The Cartesian unit cell is required for a common STEM, because the pixel positions (the electron irradiation points)  of a scanning image are on the Cartesian grid. The moiré magnification (M = d_moire / d_lattice) is determined from the following relations: M_x = 1 / | 1 – r_x / N |, M_y = 1 / | 1 – r_y / N |, where N is the closest integral number to r_x and r_y (see Fig. 1(b)).  Figures 2 (c) and (d-f) show a moiré HAADF image and moiré pseudo atomic column elemental maps by 102 cyclic acquisitions. The image width in the x direction (nd_rx) and M_xa_c are measured to be 239.1 and 69.1 nm from the image, and the M_x is derived to be 78.2 by the relation mentioned above with N = 4, r_x = d_rx / a_c = 4.053. The image height (nd_ry) is calculated to be 412 nm with the derived relation ( nd_ry = 0.5nNb_c(1 + [1 + 4/(Nn_bc)]^1/2), where n_bc is number of pixels for M_yb_c in image). The r_y is determined to be 4.029. The d_rx and d_ry is derived to be 3.73 and 6.43 nm. The total scanned area is estimated to be 9.82 x 10⁴ nm². The dose density on the sample was estimated to be 1.30 x 10¹⁰ electrons / nm² with the total analysis time = 2089 sec and the probe current = 24 pA. The equivalent dose density, if the analysis were performed by the conventional direct method, is estimated to be 1.23 x 10¹⁴ electrons / nm². Therefore, we could reduce the dose density to be < 10^-4 in this experiment. The results in Fig. 2 (d-f) show the clear atomic column elemental maps for Al, Si and O due to the low dose acquisition. We proposed and demonstrate a method to reduce dose density for the analysis of a real fragile sample. The reduced rate was a one-ten-thousandth of one by the conventional method.  

References

[1] E Okunishi et al, Microsc. Microanal. 12(S2) (2006), p.1150.

[2] Y Kondo and E Okunishi, Microscopy. 63 (5) (2014), p. 391.

Figures:

Figure 1. (a) The unit cell for STEM moiré method on 2D projected atomic site map of aquamarine (Be3Al2Si6O18: Fe2+). We must use Cartesian unit cell for STEM moiré method, since the pixel positions are on the Cartesian grid, in general scanning system. (b) In the case that N is not 1, a pixel interval is close to N*ac or N*bc. In this configuration, the dose density is reduced to significantly less than one in the case of N = 1.

Figure 2. Pseudo moiré HAADF image, conventional direct BF image and 2D moiré atomic column elemental maps of aquamarine. a: crystal structure, b: direct BF, c: simultaneously obtained 2D moiré HAADF image, d-f: pseudo moiré atomic column elemental maps of Al, Si and O.

« Back to The 16th European Microscopy Congress 2016

EMC Abstracts - https://emc-proceedings.com/abstract/application-of-moire-pseudo-atomic-column-elemental-mapping-to-electron-beam-sensitive-crystal-of-mineral/